Fabryk

A hyper-connected, multi-modal knowlege fabric • Part of the Textrynum project

Fabryk turns structured content (Markdown with YAML frontmatter) into a multi-modal knowledge store: a graph of relationships, a full-text search index, and a vector space for semantic similarity — all exposed via MCP tools and a CLI.

What it does

Content ingestion — Parse Markdown files with YAML frontmatter, extract sections, resolve sources
Knowledge graph — Build and traverse relationship graphs (prerequisites, extensions, related concepts) with petgraph
Full-text search — Tantivy-backed search with category filtering, incremental indexing, and freshness tracking
Vector/semantic search — LanceDB + fastembed for local embedding generation and similarity search
Hybrid search — Reciprocal rank fusion combining FTS and vector results
MCP server — 25+ tools for AI assistants to query the knowledge base via Model Context Protocol
CLI — Command-line interface for graph operations, search, and configuration
Auth — OAuth2 with Google provider, JWT validation, RFC 9728 discovery

Quick Start

Fabryk provides two umbrella crates so downstream projects only need one or two dependencies:

[dependencies]
# Core knowledge fabric (content, graph, search, auth, acl)
fabryk = { version = "0.2", features = ["full"] }

# MCP server tools (adds all MCP tool suites + server infrastructure)
fabryk-mcp = { version = "0.2", features = ["http"] }

For finer control, enable only the backends you need:

[dependencies]
# Just graph and FTS, no vector search
fabryk = { version = "0.2", features = ["fts-tantivy", "graph-rkyv-cache"] }

# MCP tools without HTTP transport (stdio only)
fabryk-mcp = "0.2"

Vendor-specific crates are added separately:

[dependencies]
fabryk-auth-google = "0.2"  # Google OAuth2 / JWKS
fabryk-gcp = "0.2"          # GCP credential detection

What you get

Adding fabryk to your project gives you the full Fabryk core library:

Module	Crate	What it provides
`fabryk::core`	`fabryk-core`	Shared types, traits, error handling, service management
`fabryk::auth`	`fabryk-auth`	Token validation trait, Tower auth middleware
`fabryk::acl`	`fabryk-acl`	Access control primitives
`fabryk::content`	`fabryk-content`	Markdown parsing, frontmatter extraction
`fabryk::fts`	`fabryk-fts`	Full-text search traits and types
`fabryk::graph`	`fabryk-graph`	Knowledge graph storage and traversal
`fabryk::vector`	`fabryk-vector`	Vector/semantic search traits and types

What you don't get

The following are not included in fabryk and must be added separately:

fabryk-mcp — MCP server infrastructure and tools (see fabryk-mcp)
fabryk-auth-google — Google OAuth2 / JWKS provider
fabryk-gcp — GCP credential detection utilities
fabryk-redis — Redis connection management
fabryk-cli — CLI framework

Feature flags

Feature	What it enables
`fts-tantivy`	Tantivy full-text search backend
`graph-rkyv-cache`	rkyv binary serialization for graph caching
`vector-lancedb`	LanceDB vector database backend
`vector-fastembed`	Local embedding generation via fastembed
`full`	All of the above

Without any features enabled, the search and graph crates provide only their trait definitions and types — no heavy dependencies are compiled.

Crate Map

Core (`fabryk`)

Crate	Purpose
`fabryk-core`	Shared types, traits, error handling, service management
`fabryk-auth`	Token validation trait and Tower auth middleware
`fabryk-acl`	Access control primitives
`fabryk-content`	Markdown parsing, frontmatter extraction
`fabryk-fts`	Full-text search (Tantivy backend)
`fabryk-graph`	Knowledge graph storage and traversal (petgraph)
`fabryk-vector`	Vector/semantic search (LanceDB + fastembed)

MCP (`fabryk-mcp`)

Crate	Purpose
`fabryk-mcp-core`	MCP server infrastructure, tool registry (rmcp)
`fabryk-mcp-auth`	RFC 9728/8414 OAuth2 discovery endpoints
`fabryk-mcp-content`	Content and source MCP tools
`fabryk-mcp-fts`	Full-text search MCP tools
`fabryk-mcp-graph`	Graph query MCP tools
`fabryk-mcp-semantic`	Hybrid search MCP tools

Vendor-specific (separate dependencies)

Crate	Purpose
`fabryk-auth-google`	Google OAuth2 / JWKS provider
`fabryk-gcp`	GCP credential detection
`fabryk-redis`	Redis connection management
`fabryk-cli`	CLI framework with graph commands

Key Dependencies

Component	Library	Purpose
Knowledge Graph	petgraph	Graph data structures and algorithms
Full-Text Search	Tantivy	Rust-native search engine
Vector Search	LanceDB	Embedded vector database
Embeddings	fastembed	Local embedding generation
MCP Server	rmcp	Model Context Protocol SDK
Auth	jsonwebtoken	JWT validation
CLI	clap	Command-line parsing
Configuration	confyg	Hierarchical config

License

Apache-2.0

fabryk 0.4.2